An automated speech recognizer (ASR) may use a speaker-dependent adaptation profile to increase speech recognition performance. A profile may be periodically updated, for example at the end of a user's speech session.